What problem does it solve?
This Skill converts legacy platform test runner stdout logs into ShareGPT-format JSON so downstream evaluation and iteration tools can consume conversation records consistently and automatically.
Core Features & Use Cases
- Structured Log Parsing: Parse runner node markers and map agent outputs, tool calls, tool responses, and summaries into ShareGPT conversation roles.
- Cached Converter Scripts: Detect and reuse existing scripts named log_converter_*.py to speed up repeated conversions for known formats.
- On-the-fly Parser Generation: When no cached script matches, analyze the log format, generate a Python converter that accepts --input and --output, and persist it for reuse.
- Metadata Enrichment & Robust Error Handling: Extract metadata from query.txt and config_snapshot.yaml, record timestamps, mark partial parses with parse_quality, and fall back to inline parsing on cached-script failures.
- Use Case: Integrate legacy platform test outputs into eval-judge and iteration pipelines by converting execution_log.txt files into run_log.json records.
Quick Start
Convert the directory containing execution_log.txt into a ShareGPT run_log.json using the meta-log-converter, and cache any generated parser script for reuse.
